The University of Chicago The Booth School of … · The University of Chicago ... Introduction...

The University of ChicagoThe Booth School of Business of the University of ChicagoThe University of Chicago Law School

Media versus Special InterestsAuthor(s): Alexander Dyck, David Moss, and Luigi ZingalesSource: Journal of Law and Economics, Vol. 56, No. 3 (August 2013), pp. 521-553Published by: The University of Chicago Press for The Booth School of Business of the University ofChicago and The University of Chicago Law SchoolStable URL: http://www.jstor.org/stable/10.1086/673216 .

Accessed: 04/09/2014 14:07

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Press, The University of Chicago, The Booth School of Business of the University ofChicago, The University of Chicago Law School are collaborating with JSTOR to digitize, preserve and extendaccess to Journal of Law and Economics.

http://www.jstor.org

This content downloaded from 128.100.41.75 on Thu, 4 Sep 2014 14:07:56 PMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=ucpress

http://www.jstor.org/action/showPublisher?publisherCode=chicagobooth

http://www.jstor.org/action/showPublisher?publisherCode=chicagobooth

http://www.jstor.org/action/showPublisher?publisherCode=chicagolaw

http://www.jstor.org/stable/10.1086/673216?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


521

[Journal of Law and Economics, vol. 56 (August 2013)]� 2013 by The University of Chicago. All rights reserved. 0022-2186/2013/5603-0017$10.00

Media versus Special Interests

Alexander Dyck University of Toronto

David Moss Harvard University

Luigi Zingales University of Chicago

Abstract

We argue that profit-maximizing media help to overcome the rational ignoranceproblem highlighted by Anthony Downs. By collecting news and combining itwith entertainment, media are able to inform passive voters about regulationand other public policy issues, acting as a (partial) counterbalance to small butwell-organized groups. To show the impact this information has on regulation,we document the effect muckraking magazines had on the voting patterns ofU.S. representatives and senators on regulatory issues in the early part of thetwentieth century. We also discuss the conditions under which media can serveto counterbalance special interests.

[T]here is only one way to get a democracy on its feet in thematter of its individual, its social, its municipal, its State, its Na-tional conduct, and that is by keeping the public informed aboutwhat is going on. There is not a crime, there is not a dodge, thereis not a trick, there is not a swindle, there is not a vice which doesnot live by secrecy. Get these things out in the open, describethem, attack them, ridicule them in the press, and sooner or laterpublic opinion will sweep them away. (Joseph Pulitzer)1

1. Introduction

According to the economic theory of regulation (ETR), government interventionis shaped by the competition among interest groups, where the winners are the

1 Quoted in Ireland (1914, p. 115).

We wish to thank Jonathan Lackow, Adair Morse, Marco di Maggio, and Peter Epstein for out-standing research assistance and Sam Peltzman and participants at seminars at the University ofChicago and the National Bureau of Economic Research for very useful comments. Alexander Dyckthanks the Connaught Fund of the University of Toronto, David Moss thanks the Division of Researchand Faculty Development of Harvard Business School, and Luigi Zingales thanks the Center forResearch on Security Prices, the George Stigler Center at the University of Chicago, and the Initiativeon Global Markets for financial support.



522 The Journal of LAW& ECONOMICS

groups with the lowest cost of organizing relative to the per capita benefit oforganizing (Olson 1965; Stigler 1971; Peltzman 1976). In the context of industryregulation, the ETR predicts that the winners tend to be small groups with astrong interest. Yet in practice we observe many regulations, ranging from an-titrust law to zoning restrictions, that aim to protect small and dispersed interestsagainst big and concentrated ones.

One possible explanation for the existence of regulation that seems to favordispersed groups is that such regulation may only appear to be in the interestof these groups, when in fact it secretly serves the interests of large incumbents.This explanation is consistent with the major role that large incumbents oftenplay in shaping regulatory implementation. Yet it has difficulties explaining whylawmakers enacted these types of regulation in the first place, often over thestrong objections of large incumbents. Even if some incumbents might havebenefited from antitrust law, in most instances large incumbents’ interest likelywould have been better served with no antitrust legislation at all, or at least somany incumbents believed when federal antitrust law was first enacted in thelate nineteenth century. So why do so many public policies that appear to con-strain large incumbents exist?

An alternative to the ETR is the public interest view, where regulation ariseswhenever the social benefits outweigh the social costs. Proponents of this morebenign view, however, traditionally have had trouble explaining how dispersedinterests are able to come together and solve the collective action problem,overpowering more concentrated interests where the costs of organizing are muchless.

A third explanation, which we explore in this paper, is that in a representativedemocracy, profit-maximizing media may play an important role in minimizingcollective action costs for dispersed actors and therefore may help to tip thepolitical balance against concentrated interests (including large incumbents) incertain cases. By helping to inform the broad public about certain public policyissues, media can create a more informed electorate to which politicians arelikely to try to cater. Recognizing this role that the media can play allows for aricher theory of regulation, where outcomes depend on the relative strength ofthe media in informing and, in turn, empowering dispersed versus concentratedinterests.

To understand how profit-seeking media can undermine the simple predictionsof the ETR, we need to return to its conceptual foundation. The reason whynarrow, concentrated interests dominate is that it is rational for broad, dispersedinterests not to invest in informing themselves, given the infinitesimal payoffthey receive from doing so. This is Downs’s (1957) theory of rational voterignorance. If voters remain ignorant about public policies that would serve theirinterest, it does not benefit politicians to try to protect them by enacting thesepolicies. Media, however, reduce the cost to voters (and others) of becominginformed, and they do this in two ways. First, by collecting, verifying, andsummarizing relevant facts, media minimize the collective action problem as-



Media versus Special Interests 523

sociated with gathering information for a dispersed group. Second, by repack-aging information in a way that makes it entertaining, media create a compellingrationale for individuals to bear the small cost of obtaining the gathered infor-mation. Even if it is not in each individual’s economic interest to become in-formed about a policy issue, the utility benefit provided by the entertainmentcomponent of a news story (for example, the scandalous, shocking, or titillatingdimension) can more than compensate for the costs to the individual of obtainingthe information, including the price of the newspaper and the time spent readingit. Thus, media can potentially overcome the rational ignorance problem. Butdo they have an interest in doing so?

The increasing-returns-to-scale technology used by most media induces themto cater their provision of news to the interests of large groups (Stromberg 2004b).This can be considered a negative outcome, as in Stromberg’s (2004a) view thatthis tendency generates a welfare-reducing bias toward policies that favor largegroups (Prat and Stromberg 2013). In the context of regulation, however, thisbias in profit-seeking media generates a natural counterbalance to the power ofsmall, concentrated groups, which lies at the center of traditional ETR. Why?By informing large and dispersed groups about policy-relevant issues, mediacreate an interest among politicians to cater to them.

How important is the counterbalancing role played by profit-seeking mediain shaping actual regulation? In the age of national television (not to mentionthe Internet), it is very difficult to tell.2 Since nearly everyone is exposed to newsat the same time, it is difficult to know whether media content affects voters orvoters’ demand for information drives media content.

To address this challenge, we look back in time to the muckraking era (1902–17). This period saw the rise of investigative journalists who wrote about con-temporary events with the express purpose of changing public attitudes andlegislative behavior. Because this era predates national radio and television, ex-posure to the ideas in these articles would be greatest for those citizens whoread the magazines. Since sales of these magazines differed by congressionaldistrict, we can expect the influence of their articles to vary with sales in thedistrict. We focus our attention on circulation data by congressional district andby state that we collected for two of the most prominent muckraking magazines,McClure’s and Cosmopolitan.

We analyze roll call votes of U.S. representatives on all domestic regulatorylegislation from 1902 to 1917. To control for the ideological preferences of electedrepresentatives and the role of regional factors in voting behavior, in additionto party and regional dummies, we use the x- and y-coordinates of Poole (2004),which were developed in a series of articles by Poole and Rosenthal (1997, 2001).We show that representatives voted differently on regulatory issues that were

2 Recent successful examples of how this challenge can be addressed in special circumstances areDellaVigna and Kaplan (2007), who use the geographic expansion of Fox News Channel, and Georgeand Waldfogel (2006), who exploit of the gradual expansion of national distribution of the New YorkTimes.




previously exposed in muckraking magazines, the more so the more diffusedwere muckraking magazines in their districts.

The advantage of our specification is that it controls for any specific char-acteristics of the representatives and the district, and we focus our attention onthe results based on the interaction between issues that are muckraked anddiffusion of muckraking magazines. The disadvantage is that we cannot rule outthe possibility that voters who are more sensitive to certain issues are more likelyto buy magazines that cover those issues and thus that the political preferencesof voters in a particular district may be the source of both the voting behaviorof their congressional representatives and the diffusion of muckraking magazinesin their districts.

To try to address this concern, we test the same specification on the votingbehavior of senators to see if exposure to muckraking and the interaction ofexposure and diffusion influenced their voting behavior. Until 1913 the vastmajority of senators were appointed, not elected. This process ensured thatsenators reflected the ideology or preferences of the governor or the state leg-islature that appointed them, but not necessarily the sensitivities or preferencesthat voters developed after becoming informed about particular issues. If muck-raking magazine circulation is just a proxy for the prevailing ideology in a certainarea, then this circulation should affect the way senators voted as well. It doesnot, in general. We do find some impact, however, if we split the sample betweensenators who were in states that provided for direct election and senators whowere not directly elected. Directly elected senators are affected, much like theirelected counterparts in the House of Representatives, whereas appointed senatorsare not.

This test alleviates the concern that our results are driven solely by the factthat muckraking magazine circulation is a proxy for the prevailing ideology ina certain area. Yet it is still possible—albeit unlikely—that our results are drivenby the fact that muckraking magazine circulation is a proxy for some voters’preferences that are not captured by a senator’s party affiliation or by his Pooleand Rosenthal coordinates.

Taking this Senate and House evidence together illustrates the mechanismthrough which profit-seeking media influence regulatory outcomes. Elected of-ficials are sensitive to informed voters, and media help voters become informedeven on issues where their direct interest in being informed is minimal. Thispoint is different from, but complementary to, that made in Stromberg (2004a,2004b). Stromberg shows that media are effective in informing interested votersof the policies that benefit them directly. We show here that the point is moregeneral and also applies to public interest policies in which few, if any, votershave a large economic interest.

To illustrate this point further, we examine the influence of media reportingon voting behavior around a reform that, while arguably benefiting the countryoverall, was not in the specific interest of any group: the Seventeenth Amendment,which provided for direct election of senators. Prior to ratification of this amend-




ment, most senators were appointed rather than elected. The insulation of thesenate selection process from the will of the voters allowed business interests toexert considerable control over senate appointments and to pick sympatheticsenators who were likely to support their special interests. In 1906, one of thelargest muckraking magazines, Cosmopolitan, published David Graham Phillips’Treason of the Senate series (Phillips 1906a–1906i). The series stirred enormouscontroversy, accusing the Senate of being “the eager, resourceful, and indefati-gable agent of interests as hostile to the American people as any invading armycould be” (Phillips 1906g, p. 488). Did this exposure of weakness—including,potentially, corruption—in the appointment process influence senators’ votingbehavior?

To test this idea, we examine patterns in Senate votes on the SeventeenthAmendment that preceded (in 1902) and followed (in 1911) the publication ofthis series. These votes provide us with a quasi experiment with which to studythe effect of exposure to the Treason of the Senate series on senators’ votingbehavior. Senators from states where Cosmopolitan was more highly diffusedwere more likely to switch their votes (from negative in 1902 to positive in 1911).This effect cannot be explained in terms of the state’s electorate having a highersensitivity toward the issue, since we show that the senator from the same statevoted differently before the issue was muckraked. Similarly, this effect is notsimply that areas where muckraking magazines were more diffused became moresympathetic to the idea of an elected Senate, since the diffusion of anothermuckraking magazine, McClure’s (which did not feature the Treason of the Senateseries), had no independent effect on voting behavior.

In showing that media during the muckraking period contributed to regulationthat favored broad and dispersed groups, we are not suggesting that this is alwaysthe case, nor that the outcome (in this case or any other) is necessarily optimalfrom a social welfare standpoint. Media involvement may be insufficient toovercome the power of large incumbents and other concentrated interests, orits overwhelming effect may be so strong as to trigger impulsive and potentiallyinefficient initiatives. Rather, the point we want to emphasize is that these resultssuggest that the media can play a countervailing role to the power of concentratedinterests in the theory of regulation. To transform this insight into a positivetheory of regulation, we need to address the question of when media are moreor less likely to play this countervailing role. We discuss this issue in Section 5.

Our analysis of the impact of muckraking is related to a growing literatureon the effect of media coverage on a range of political factors, including votingbehavior (DellaVigna and Kaplan 2007), government intervention (Besley andBurgess 2002; Moss and Oey 2010), subsidies (Stromberg 2004b), and foreignaid (Eisensee and Stromberg 2007).3 Like Stromberg (2004a, 2004b), we usehistorical evidence to identify differential exposure to news. In showing thatmuckraking facilitated the approval of Progressive Era legislation, our paper is

3 For a nice summary of this work, see DellaVigna and Kaplan (2008).




similar to Law and Libecap (2006), which documents that vested interests haveless power in explaining congressional votes on the Pure Food and Drug Actafter the publication of muckraking stories. We add to Law and Libecap (2006)in two ways. First, we provide a mechanism through which media in general(and not just the muckrakers) help counteract the power of vested interests.Second, through a combination of time-series and cross-sectional evidence ona range of regulatory legislation, we are able to identify the working of thismechanism.

The remainder of the paper proceeds as follows. Section 2 discusses how mediacan overcome rational ignorance by entertaining their customers in the contextof policy-relevant issues. Section 3 tests the impact of media information onpolicy outcomes by focusing on representatives’ and senators’ voting behavioron regulatory issues and how it is influenced by muckraking and the diffusionof muckraking magazines. Section 4 tests the impact of one of the major muck-raking series, Treason of the Senate, on the voting behavior of U.S. senators onthe Seventeenth Amendment. Section 5 discusses the conditions under whichthe media are more likely to act as a countervailing force to private specialinterests. Section 6 concludes.

2. The Role of the Media in Overcoming Rational Ignorance

Starting with Stigler (1971), the economic theory of regulation presupposesthat voters remain poorly informed about regulatory issues. The foundation forthis assumption goes back to Downs (1957), who suggests that it was rationalfor voters not to invest in acquiring such information on their own since thepayoff (in terms of influence over policy outcomes) was infinitesimally small forindividual voters. There are two cost-based explanations for why this result holds.First, there is a collective action problem in gathering the information. Whileeveryone might benefit from such information being gathered, no one individ-ually wishes to cover the cost of collecting it. Second, even if the collective actionproblem were resolved and a third party took charge of collecting, verifying, andsummarizing the information, both the per capita cost of the information andthe individual cost of processing it (for example, the time involved in readingor viewing the news) might still exceed the payoff in terms of the marginallyincreased likelihood of a more favorable policy outcome as a result of a moreinformed vote being cast.4

The media can potentially resolve both of these problems. First, by collecting,verifying, and summarizing relevant facts, the media can essentially resolve thecollective action problem. In fact, each media outlet may be interpreted as anagent delegated by the multitude of its customers to collect information on their

4 Naturally, in practice there may be other reasons for individuals to become informed aboutpolicy-relevant issues and facts (including civic mindedness), which the economic theory of regulationtends to overlook.




behalf. This delegation of responsibility (and pooling of resources) solves thecollective action problem but simultaneously introduces an agency problem—namely, ascertaining whose interest determines exactly what information iscollected.

Second, by repackaging information in a way that makes it entertaining, themedia may succeed in inducing voters to process it and thus become informed.Even if the cost of processing the information remains greater for the voter thanthe expected benefit to be derived from a more informed vote, the utility providedby the entertainment component may repay readers for both the cost (if any)of obtaining the information from the media outlet (namely, the price of thenewspaper) and the time spent absorbing the information, thus making it worth-while to become informed.

That people may become informed for consumption reasons was an ideaalready present in Downs (1957) and emphasized by Hamilton (2004), but ithas not played a big role in the economic literature as a factor in overcomingrational apathy. In fact, Becker and Murphy (1993) emphasize the oppositechannel: advertisers pay for entertainment to inform their customers. In ourcase, information is a simple by-product of the production of entertainment, asin the program The Daily Show featuring Jon Stewart.

An alternative way to overcome voters’ rational apathy is to assume a privatereturn from becoming informed; for example, one may learn how best to exploitsubsidies (Stromberg 2004b). While this aspect is important in many types ofwelfare legislation, it is unlikely to be the primary force in the case of regulatorydecisions, such as the Clean Air Act, or constitutional amendments, such as theSeventeenth Amendment. For the public to become informed about these sortsof issues (where there is little private benefit), the entertainment factor is a morelikely facilitating mechanism. Once voters become informed, several studies(Baron 1994; Grossman and Helpman 1996; Stromberg 2004a) show theoreticallywhy it is harder for elected representatives to cater to special interests.

This link between media coverage and regulatory outcomes is difficult to testpersuasively in an age of national television, not to mention the Internet. Thanksto these modern media, nearly everyone can potentially be exposed to the samenews at about the same time, which makes the identification problem verychallenging: how can we identify the causal link only from the time series, whenmany other events occurred at the same time?

3. Evidence from the Muckraking of Regulatory Issues

Before the introduction of national radio and television broadcasting, therewas great diversity in exposure to news. For this reason, to identify whether presscoverage influences the political support for regulation, we focus on the muck-raking period of American journalism in the early years of the twentieth century.Changes in technology and demand created fertile ground for an explosion of




newspapers and magazines, such as Cosmopolitan, Everybody’s, McClure’s, andCollier’s (see Hamilton 2004; Glaeser and Goldin 2006). These magazines soughtand realized a mass audience. Conservative estimates place sales of all suchmagazines at 3 million, the top four magazines at 2 million, and total readershipas high as 20 million, all at a time when the U.S. population was 80–90 millionand 15 million people voted in the national presidential election (Fitzpatrick1994, p. 108).5

Initially focusing mainly on fictional accounts, these magazines reached a farbroader audience by covering real-world scandals of all sorts. A quick perusalof the titles of notable articles reveals their focus. In 1905 and 1906, attentionfocused on producers of medicines and meat-packers in articles such as thosein the Great American Fraud series (Adams 1905a–1905d, 1906a–1906f), “ThePatent Medicine Conspiracy against Freedom of the Press” (Sullivan 1905), “IsChicago Meat Clean” (Sinclair and Seaman 1905), “Stockyard Secrets” (Sinclair1906b), and “The Condemned-Meat Industry” (Sinclair 1906a). Others, such as“Water Power and the Price of Bread” (Mathews 1908) and “Water Power andthe Pork Barrel” (Mathews 1909) were featured in 1908 and 1909, and throughoutthe most active muckraking period (1902–12), there appeared numerous inves-tigations of business power and political corruption, such as the Treason of theSenate series in 1906 and “An Exposition of the Sovereign Political Power ofOrganized Business” (Steffens 1910, 1911).

Because these magazines were not equally read throughout the country, wecan use data on their coverage and circulation to test the models of Baron (1994),Grossman and Helpman (1996), and Stromberg (2004a) that show that thefraction of informed voters alters the balance of power between private interestand public interest.6

Before discussing our tests, however, we need to explain how we collectedthese data, identified those pieces of legislation that were muckraked and thoseareas of the country that were more exposed to muckraking ideas, and codedlegislators’ voting behavior.

3.1. Sample of Regulatory Legislation

To assemble a list of all regulatory legislation with available roll call votingrecords, we start with the Voteview data set (Poole 2004) and use Poole’s clas-sification of the votes. Following Peltzman (1976), Poole classified all votes intoone of eight categories. We focus on the two categories of regulation (regulationgeneral interest and regulation special interest), where all of the literature, startingwith Stigler (1971), suggests that special interests will be most active. We assem-

5 The numbers of votes cast in presidential elections were 13.52 million in 1904, 14.88 million in1908, 15.04 million in 1912, and 18.53 million in 1916 (U.S. Bureau of the Census 1975).

6 We are not suggesting that these muckraking articles were not reproduced or conveyed elsewhere,as Law and Libecap (2006) illustrate for the debate surrounding patent medicine legislation. Rather,we are relying on the assumption that exposure to these ideas was more intense in areas with highercirculation.




bled all such regulation votes from the Fifty-Seventh through Sixty-Fourth U.S.Congresses, both House and Senate (1902–17), which includes and slightly ex-tends the period generally understood to be the era of muckraking (Weinbergand Weinberg 1961; DeNevi and Friend 1973).

To make our task more manageable, we further restrict ourselves to final votes7

and votes on issues that relate to domestic policy.8 When votes on the same billoccurred in both legislative chambers, we included both votes, even if only onewas classified as a vote on regulation.9

The final sample of legislation is provided in Table 1, which includes 40 finalvotes in the House and 34 final votes in the Senate. This list includes almost allof the notable legislation associated with muckraking, such as that which createdthe antecedent to the modern U.S. Food and Drug Administration. It does notinclude the Seventeenth Amendment, which we evaluate later since it is notclassified as regulation.

3.2. Issues Covered in Muckraking Magazines

To measure the coverage of issues that are relevant to legislation, we examinedall of the famous muckraking articles. We started with the book The Muckrakers(Weinberg and Weinberg 1961), which categorizes and reprints 27 notable muck-raking articles and includes an uncategorized bibliography listing 98 importantbut less notable muckraking articles.10

After reading the notable articles and reviewing the less notable ones, weconstructed a measure of whether a regulatory issue was muckraked. We assignthe term Muckraked a value of one if the regulatory issue was covered in muck-raking magazines and zero otherwise. As Table 1 shows, 23 of 40 House votes

7 We searched using the keywords “Pass S” and “Pass H,” which retrieve only the votes whosedescription indicates a vote to pass a bill, marked by “H.R.” or “S.”

8 A number of votes dealt with managing federal territories such as Alaska, the Philippines, or theDistrict of Columbia.

9 Because of inconsistencies in coding, a vote that is classified as regulation in one chamber mightnot be coded that way in the other (Keith Poole, professor emeritus of political science, Universityof California, San Diego, e-mail to Jonathan Lackow, research assistant, June 26, 2004).

10 The categorization by Weinberg and Weinberg (1961) appears in their table of contents and isas follows: “Behind Political Doors,” which we label government corruption, and which includes thesubcategories “The United States Senate,” “The United States House of Representatives,” “The State,”“The City,” “The Ward,” and “Bureaucracy”; “Poison—Beware,” which we label food and drug, andwhich includes the subcategories “Patent Medicine” and “Pure Food”; “People in Bondage,” whichwe do not use, and which covers racial issues; “High Finance,” which includes the subcategories“Mother of Trusts,” which we label monopoly; “Stock Market,” “Insurance,” and “Railroads,” towhich we add the further subcategory of water power because of the preponderance of articles onwater power; “The Church,” which we do not use; “Prisons,” for which we focus on prison laborand label as such; “Labor,” including the categories “Workmen’s Compensation” and “Child Labor,”and to which we add a further subcategory, working hours; and, finally, “Vice,” for which we focuson liquor and label as such.



530

Tab

le1

Dom

esti

cR

egu

lato

ryLe

gisl

atio

nw

ith

Fin

alR

oll

Cal

lV

otes

,19

02–1

7

Dat

eC

ham

ber

Subj

ect

Mu

ckra

ked

Jan

uar

y31

,19

02H

ouse

Subj

ect

oleo

mar

gari

ne

tost

ate

law

s0

Mar

ch17

,19

02Se

nat

eO

cean

mai

lse

rvic

ean

dde

ep-s

eafi

sher

ies

0A

pril

3,19

02Se

nat

eTa

x,re

gula

te,

and

defi

ne

oleo

mar

gari

ne

(pas

s)0

Jan

uar

y14

,19

03H

ouse

Reb

ate

duti

eson

coal

0Ja

nu

ary

14,

1903

Hou

seR

ebat

edu

ties

onco

al0

Febr

uar

y7,

1903

Hou

seP

rote

ctco

mm

erce

agai

nst

mon

opol

ies

1Fe

bru

ary

13,

1903

Hou

seR

egu

late

inte

rsta

teco

mm

erce

1M

arch

2,19

03H

ouse

Law

sco

nce

rnin

gto

bacc

o0

Mar

ch3,

1903

Hou

seG

ives

pref

eren

ceto

vete

ran

sfo

rci

vil

appo

intm

ents

0M

arch

1,19

04Se

nat

eR

equ

ire

that

carg

oes

carr

ied

onU

.S.

vess

els

befo

rpu

blic

purp

oses

0Fe

bru

ary

6,19

05H

ouse

Reg

ula

tera

ilroa

dra

tes

1Fe

bru

ary

9,19

05H

ouse

Reg

ula

tera

ilroa

dra

tes

1Fe

bru

ary

21,

1906

Sen

ate

Pre

ven

tm

isbr

andi

ng

offo

od,

dru

gs,

and

liqu

or(p

ass)

1A

pril

2,19

06H

ouse

Law

sab

out

fort

ifica

tion

ofpu

resw

eet

win

es0

Apr

il2,

1906

Hou

seLa

ws

abou

tfo

rtifi

cati

onof

pure

swee

tw

ines

0Ju

ne

23,

1906

Hou

seP

ure

food

and

dru

gs1

Jan

uar

y10

,19

07Se

nat

eP

rom

ote

safe

tyon

rails

and

limit

serv

ice

hou

rsof

empl

oyee

s(p

ass)

1Fe

bru

ary

18,

1907

Hou

seLi

mit

railr

oad

empl

oyee

s’h

ours

ofse

rvic

e0

Apr

il6,

1908

Hou

seLi

abili

tyof

com

mon

carr

iers

1A

pril

21,

1908

Hou

seSi

mpl

ify

cust

oms

law

s0

Apr

il24

,19

08H

ouse

Safe

tydu

rin

gre

gatt

asan

dm

arin

eev

ents

0M

ay13

,19

08H

ouse

Pro

tect

pate

nts

0



531

May

13,

1908

Hou

seP

reve

nt

impo

rtof

impu

rete

a0

May

14,

1908

Hou

seP

rote

ctba

nk

depo

sito

rs0

May

15,

1908

Sen

ate

Am

end

nat

ion

alba

nk

law

s(p

ass)

0M

ay26

,19

08H

ouse

Reg

ula

teex

plos

ives

inin

ters

tate

com

mer

ce0

May

27,

1908

Hou

seR

egu

late

the

max

imu

mn

um

ber

ofst

eera

gepa

ssen

gers

0Ju

ly8,

1909

Sen

ate

Pro

vide

gove

rnm

ent

reve

nu

e,eq

ual

ize

duty

,an

den

cou

rage

U.S

.in

dust

ry(p

ass)

1Ju

ne

3,19

10Se

nat

eR

egu

late

inte

rsta

teco

mm

erce

(pas

s)1

Febr

uar

y15

,19

11Se

nat

eP

rote

ctw

ater

shed

san

dap

poin

tco

mm

itte

eto

acqu

ire

lan

d1

May

8,19

11H

ouse

Pla

ceag

ricu

ltu

ral

impl

emen

tson

free

list

1Ju

ly27

,19

11Se

nat

eR

edu

cew

ool

duty

(pas

s)1

July

27,

1911

Sen

ate

Red

uce

woo

ldu

ty(p

ass)

1A

ugu

st1,

1911

Sen

ate

Pla

ceva

riou

sit

ems

onfr

eelis

t(p

ass)

1A

ugu

st1,

1911

Sen

ate

Pla

ceva

riou

sit

ems

onfr

eelis

t(p

ass)

1A

ugu

st3,

1911

Hou

seR

edu

cedu

tyon

cott

ongo

ods

1A

ugu

st17

,19

11Se

nat

eR

edu

ceco

tton

man

ufa

ctu

redu

ty(p

ass)

1Ja

nu

ary

29,

1912

Hou

seE

qual

ize

duti

es1

Febr

uar

y21

,19

12H

ouse

Equ

aliz

edu

ties

1M

arch

15,

1912

Hou

seE

qual

ize

duti

es1

Mar

ch19

,19

12H

ouse

Ext

end

spec

ial

exci

seta

x0

Apr

il1,

1912

Hou

seR

edu

cedu

ties

onw

ool

and

woo

len

good

s1

May

6,19

12Se

nat

eP

rovi

deco

mpe

nsa

tion

for

inju

ryto

orde

ath

ofco

mm

on-c

arri

erem

ploy

ees

(pas

s)1

May

30,

1912

Sen

ate

Pro

vide

gove

rnm

ent

reve

nu

e,eq

ual

ize

duty

(pas

s)1

May

31,

1912

Sen

ate

Lim

ith

ours

ofla

bore

rsan

dm

echa

nic

s1

July

25,

1912

Sen

ate

Red

uce

duty

onw

ool

and

woo

lpr

odu

cts

(pas

s)1

July

26,

1912

Sen

ate

Ext

end

spec

ial

exci

seta

x(p

ass)

0Ju

ly27

,19

12Se

nat

eP

rovi

dere

ven

ue,

equ

aliz

edu

ties

,an

den

cou

rage

indu

stry

1A

ugu

st2,

1912

Hou

seD

uty

onco

tton

man

ufa

ctu

res

1A

ugu

st14

,19

12Se

nat

eR

edu

cedu

tyon

cott

onm

anu

fact

ure

s(p

ass)

1Ja

nu

ary

11,

1913

Sen

ate

Pro

hibi

tin

ters

tate

com

mer

ceof

liqu

or(p

roce

dura

l)1

Febr

uar

y8,

1913

Hou

seIn

ters

tate

ship

men

tof

liqu

or1

Mar

ch1,

1913

Hou

seW

orkm

an’s

com

pen

sati

onin

Was

hin

gton

,D

.C.,

and

inte

rsta

teco

mm

erce

1Se

ptem

ber

9,19

13Se

nat

eP

rovi

dego

vern

men

tre

ven

ue,

equ

aliz

edu

ties

(pas

s)1

Febr

uar

y2,

1914

Sen

ate

Inte

rsta

teco

mm

erce

com

mit

tee

advi

ses

gove

rnm

ent

toin

stit

ute

suit

s(p

ass)

1M

arch

4,19

14H

ouse

Reg

ula

teco

nvi

ct-l

abor

-mad

ego

ods

1



Tabl

e1

(Con

tinu

ed)

Dat

eC

ham

ber

Subj

ect

Mu

ckra

ked

May

1,19

14Se

nat

eIn

spec

tgr

ain

inin

ters

tate

com

mer

ce(p

ass)

1Ju

ne

5,19

14H

ouse

Reg

ula

tein

ters

tate

com

mer

ce1

Au

gust

4,19

14H

ouse

Reg

ula

teco

nst

ruct

ion

ofda

ms

1A

ugu

st5,

1914

Sen

ate

Cre

ate

inte

rsta

tetr

ade

com

mit

tee,

defi

ne

pow

er(p

ass)

1Se

ptem

ber

2,19

14Se

nat

eA

nti

tru

stre

gula

tion

s(p

ass)

1Ja

nu

ary

4,19

15H

ouse

Un

ifor

mgr

ain

grad

ing

1Ja

nu

ary

13,

1915

Hou

seD

efin

est

anda

rdba

rrel

for

fru

its

and

vege

tabl

es,

incl

udi

ng

for

expo

rtat

ion

0Fe

bru

ary

2,19

16H

ouse

Pro

hib

itch

ildla

bor

item

sfr

omin

ters

tate

com

mer

ce1

Febr

uar

y16

,19

16Se

nat

eC

omm

itte

efo

rin

ters

tate

com

mer

ce(p

ass)

1M

arch

8,19

16Se

nat

eR

egu

late

dam

con

stru

ctio

n(p

ass)

1M

ay20

,19

16H

ouse

Dev

elop

mer

chan

tm

arin

efo

rU

.S.

com

mer

ce1

Au

gust

8,19

16Se

nat

eP

reve

nt

inte

rsta

teco

mm

erce

ofch

ildla

bor

prod

uct

s(p

ass)

1A

ugu

st18

,19

16Se

nat

eM

erch

ant

mar

ine

bill

(pas

s)1

Sept

embe

r1,

1916

Hou

seE

stab

lish

8-h

our

wor

kday

for

peop

leem

ploy

edin

inte

rsta

teco

mm

erce

1Se

ptem

ber

2,19

16Se

nat

eE

stab

lish

8-ho

ur

wor

kday

for

carr

ier

empl

oyee

s1

Sept

embe

r5,

1916

Sen

ate

Incr

ease

reve

nu

ere

gard

ing

duti

es(p

ass)

1Se

ptem

ber

5,19

16Se

nat

eR

epor

tre

gard

ing

tari

ffto

prot

ect

indu

stri

esan

dpr

even

tm

onop

olie

s0

Jan

uar

y4,

1917

Hou

seW

ater

may

bedi

vert

edfr

omN

iaga

raFa

lls1

Not

e.T

he

sam

ple

incl

ude

sal

lvo

tes

inth

eV

otev

iew

data

set

(Poo

le20

04)

labe

led

asre

gula

tion

(reg

ula

tion

gen

eral

inte

rest

and

regu

lati

onsp

ecia

lin

tere

st).

The

vari

able

Mu

ckra

ked

take

sth

eva

lue

ofon

eif

the

issu

ece

ntr

alto

the

vote

was

feat

ure

din

mu

ckra

kin

gm

agaz

ine

arti

cles

.




Figure 1. Diffusion of McClure’s per 1,000 people in 1915

were subject to discussion in the muckraking magazines, while 28 of 34 Senatevotes were muckraked.11

3.3. Cross-Sectional Differences in Exposure to Muckraking Magazines:Circulation by District

We also construct a measure of cross-district differences in exposure to articlesin muckraking magazines. Here we exploit the fact that we have been able toassemble detailed data on circulation by city and town for McClure’s, one of themost prominent muckraking magazines of that era, which had circulation num-bers ranging from 360,000 to more than 500,000 per issue during this period.In 1917 McClure’s published a detailed breakdown of its circulation, providingcirculation not only by state but also for every town with a population greaterthan 5,000 people (McClure Publications 1917). Figure 1 provides a first glimpseat the heterogeneity in circulation, aggregated to the state level to illustrate cross-state differences. McClure’s circulation was not highly correlated with urbani-zation rates (.34), and it was even less correlated with newspaper circulation percapita (.11).

Table 2 provides summary statistics on the circulation that is the focus of ourtests, which is the more finely disaggregated circulation per district, which av-eraged 1,330 per district. At that time, the average size of an electoral district

11 In Dyck, Moss, and Zingales (2008), we introduced, in addition to a one/zero coding of whetheran article was muckraked, a subjectively coded measure of the intensity of muckraking. Since ourresults were qualitatively and quantitatively similar with both measures, here we use the more objectivemuckraking dummy.




Table 2

Summary Statistics for Votes on Domestic Regulatory Legislation, 1902–17

Mean SD Min Max

House regulatory vote regressions (N p 7,497):Pro-consumer vote .47 .499 0 1x-Coordinate .028 .478 �.905 .987y-Coordinate �.030 .512 �1.268 1.342Republican .473 .499 0 1Third party .008 .088 0 1McClure’s Circulation (district) 1,330 1,187 90 11,284McClure’s Circulation % .689 .614 .047 5.84Muckraked .634 .482 0 1McClure’s Circulation % # Muckraked #100 .432 .581 0 5.841

Senate regulatory vote regressions (N p 2,161):Pro-consumer vote .523 .499 0 1x-Coordinate �.001 .435 �1.22 .821y-Coordinate �.227 .6389 �1.381 1.386Republican .518 .500 0 1Third party .003 .053 0 1McClure’s Circulation (state) 11,265 13,681 1,106 78,242McClure’s Circulation (state)/1910 population .651 .357 .162 1.526Muckraked .805 .396 0 1McClure’s Circulation % # Muckraked # 100 .526 .414 0 1.526

Note. The sample for the House is based on 40 final roll call votes; the sample for the Senate is basedon 34 roll call votes. The Poole and Rosenthal (1997, 2001) x- and y-coordinates represent a measure ofa representative’s ideology and geography, respectively, based on his voting record to that point. McClure’sCirculation % is the circulation of McClure’s magazine expressed as a percentage of the average district(House sample) or state (Senate sample) population. The dummy Muckraked equals one when an issuehas been covered in one of the muckraking magazines.

was 193,200 people, so on average slightly less than 1 percent of the inhabitantsbought McClure’s. This figure, though, vastly understates the potential impor-tance of muckraking magazines such as McClure’s.

On one hand, Fitzpatrick (1994) estimates that readership of muckrakingmagazines was up to six times their circulation. On the other hand, only 15percent of the population voted, and people who read McClure’s were probablymore likely to vote.

Importantly, there is significant cross-district variation in circulation, with aminimum level of 90 sales per district (North Dakota, second district), 584 salesat the twenty-fifth percentile (Pennsylvania, twelfth district), 991 at the median(Maine, third district), 1,641 sales at the seventy-fifth percentile (Ohio, eighthdistrict), and a maximum level of 11,284 sales (California, seventh district). Whilethe highest sales were in the West, they were also present in the South (593 salesin the South Atlantic, 458 in the East South Central, and 825 in the West SouthCentral).

To construct the district-level circulation we aggregate circulation by townand county by using additional information on the geographic boundaries ofdistricts from Maris (1982) (by county and sometimes town, and even by specificcity blocks where necessary). Because the number of districts and the boundaries




of the districts changed for each Congress, we recalculate the McClure’s circu-lation per congressional district for each Congress in our sample period.12 Forsome of our tests, we divide this number by the average population per districtto express circulation as a percentage of average district population.13

3.4. Voting Behavior

We seek to test whether the public attention generated by muckraking forcedrepresentatives to vote more frequently in favor of regulation that served theinterest of broad dispersed groups. This is not straightforward, since there is noobvious, nonarbitrary way to determine where the general interest lies withrespect to each piece of legislation.

To address this problem, we begin by assuming that pieces of legislationsupported by a higher proportion of Democrats than Republicans are more likelyto be in the interest of consumers. Thus, we label as consumer oriented anylegislation where the percentage of Democrats voting in favor exceeds the per-centage of Republicans voting in favor by at least 10 percentage points. If thedifference between the two percentages is less than 10 percentage points, wedrop this vote from the sample (11 House votes and four Senate votes weredropped as a result). If the percentage of Republicans voting in favor exceedsthe percentage of Democrats voting in favor by at least 10 percentage points,we label the legislation as not consumer oriented. Of the 30 remaining piecesof legislation, 12 are labeled as consumer oriented. We then classify each rep-resentative’s vote on each piece of legislation as consumer oriented if he votedin favor of a pro-consumer legislation or against an anti-consumer proposal. AsTable 2 shows, 47 percent of the votes in the House are pro-consumer. Ourmain test is whether having constituents who are exposed to muckraking pos-itively affects the probability that a representative will cast a pro-consumer vote.14

12 The source for geographic boundaries of districts is Maris (1982).13 When a town was included in our McClure’s list but was not mentioned specifically in Maris

(1982), we allocated the town to the district that in our judgment (based on maps of the area)seemed most likely. When there were multiple districts in the same city identified by McClure’s (forexample, Manhattan), we divided the city’s circulation equally across all districts in that city. Inaddition, McClure’s always had some excess circulation by state that was not attributed to specifictowns or cities (likely arising from circulation in towns of fewer than 5,000 or perhaps mismeasure-ment in the geographic breakdown). We distributed this excess circulation equally across all districtsin the state. When there was a district but no other indications of circulation, the district was assumedto have the average excess circulation. At this time, in some states there were representatives withoutcongressional districts—that is, general ticket or at-large representatives. For these cases, we attributedthe average circulation per district in the state. Finally, we restricted our sample for the Sixty-FourthU.S. Congress to those states that did not have redistricting that changed the number of representativesor the apportionment of them across general ticket or at-large and geographically assigned districts,which reduced our sample by eight states.

14 Although there was certainly nothing like a direct one-to-one connection between party affiliationand consumer orientation during the Progressive Era (in fact, many of the most notable reformersduring this period were midwestern Republicans, often characterized as insurgent Republicans), therelevant historical literature suggests that Democrats in Congress tended to vote more regularly forreform measures than did Republicans, especially after 1910. Romero (2007, p. 822) reports, forexample, that “[b]ackground literature suggests that the institutional characteristic most likely to




To address the arbitrariness of this classification in a previous version of thispaper,15 we tested instead whether muckraking led representatives to vote dif-ferently from how they normally voted,16 and we obtained very similar results.In addition, in an unreported regression, we classify as pro-consumer votes onlythose where a majority of northern Democrats were in favor. The results aresimilar.

Besides the exposure to muckraking, as determinants of pro-consumer voteswe also use the x- and y-coordinates developed by Poole and Rosenthal (1997).They gathered all the data on voting behavior in roll call votes in the U.S.Congress. On the basis of these data, they identified two factors that predictvotes, which they call the x-coordinate (ideology) and the y-coordinate (geog-raphy). The value of these variables is summarized in Table 2, which presents

have influenced reform support [during the Progressive Era] is the partisan makeup of the chamber,indicating in particular the significance of Democrats and insurgent Republicans.” Similarly, incharacterizing partisan orientation in Congress on reform legislation during the Progressive Era,Allen and Clubb (1974, p. 133) write, “Democrats tended to support progressive reform legislationmost consistently, and most Republicans opposed these same reform measures.” See also McDonagh(1992) and Sanders (1999). However, because the party-based classification system used here is farfrom perfect (and ends up filtering out some famous examples of consumer protection legislation),we also employed an alternative method in an earlier version of the paper (Dyck, Moss, and Zingales2008), which did not rely on partisan margins of victory in classifying regulatory legislation asconsumer oriented or not consumer oriented.

15 See Dyck, Moss, and Zingales (2008, esp. pp. 13–14): “We seek to test whether the publicattention generated by muckraking forced representatives to vote more frequently in favor of reg-ulation that served the general interest. This is not straightforward, since it is arbitrary (and ideo-logically charged) to determine where the general interest lies in each piece of legislation. For thisreason, we choose instead to test whether muckraking led representatives to vote differently fromwhat they normally did. To compute this deviation, we exploit the fact that political scientists havealready developed measures that they claim capture the ‘normal’ voting behavior of representatives.Our technique is simply to compare a measure of the actual voting behavior on a specific issue witha measure of their predicted normal voting behavior from these studies. Our conjecture is that therewill be greater distance between actual and predicted normal values on issues that are muckraked,as compared to those that are not. Or, stated differently, we expect the exposure provided bymuckraking to move representatives away from their traditional voting stance. As the measure ofpredicted voting behavior we use the score for the x co-ordinate developed by Keith Poole andHoward Rosenthal. . . . Poole and Rosenthal scores are unique to individual lawmakers. To makethis measure vary by issue as well, in each roll call vote we assign to each representative the averagescore of representatives who voted in the same way. As an example, suppose that there are 40Democrats and each had an x score of �0.19 and there are 60 Republicans and each had a scoreof 0.27. On a particular issue 11 Republicans join the Democrats in voting in favor. In this case thescore of all those who voted in favor would be �.091[ p ((40*(�0.19)) � (11*0.27))/51], whilethe score of those voting against would be 0.27 since all have a score of 0.27. This scoring systemhas the defect of underestimating variation in voting behavior since those who deviate in their votetend to pull the average toward them. Since this biases against finding our results, we are willing totolerate the defect. . . . In Table 3 we regress each individual vote in all ‘regulatory’ bills on arepresentative fixed effect and an indicator variable for issues that were muckraked. . . . In an issuethat was actively muckraked (muckraking variable equal to 1), a representative vote moves ‘to theleft’ by an amount equal to 73% of the mean value of the x-coordinate.”

16 To compute this deviation, we exploit the fact that political scientists have already developedmeasures that they claim capture the normal voting behavior of representatives. The results (Dyck,Moss, and Zingales 2008, table 3) are very consistent.




summary statistics for the House vote regressions in Table 3 and for the Senatevote regressions in Table 4.

3.5. Results for House Votes on Regulatory Legislation

In Table 3 we focus on House votes on regulatory legislation. We estimate aprobit model of the probability that representative j will cast a pro-consumervote on issue k:

Prob{V p 1} p f{X , X , m ),jk j k jk

where Xj and Xk are representative-specific and issue-specific control variables(including the x- and y-coordinates) and is the exposure to muckraking ofmjk

district j in issue k. The coefficients reported are the marginal effect computedat the average value of the independent variables.

Column 1 presents the basic specification, with the level of sales of McClure’sin the district standardized by population, the Poole and Rosenthal x- and y-coordinates, dummy variables for the Republican Party and the third party, anda dummy variable equal to one if an issue was muckraked. In addition, weinclude nine census division fixed effects (coefficients not reported). The standarderrors are clustered at the congressional district level.

Not surprisingly, the x-coordinate is highly significant. A higher value for thex-coordinate, which ostensibly corresponds to a more right-wing ideologicalposition, decreases the probability of a pro-consumer vote. A higher value forthe y-coordinate, which captures geography, increases the probability of a pro-consumer vote, but the effect is marginally significant. The Republican Partyand third-party dummies are not statistically significant, which is not surprisingsince we control for the x-coordinate.

As expected, the muckraking dummy has a positive and statistically significanteffect on the probability of a pro-consumer vote (a 3-percentage-point increase,equal to a 6 percent increase). This positive correlation, noted by several his-torians, lies at the heart of the existing belief that muckraking contributed toProgressive Era legislation. It does not, however, address the question of causality.It could be that muckrakers focused on topics where the sensitivity was alreadyhigh and where a pro-consumer vote was more likely, regardless of what theypublished.

More compelling (and more direct) evidence of the impact of muckrakingwould be to show that this effect is stronger for congressmen elected in districtswhere muckraking magazines were more highly diffused. This is what we showin column 2, where we include the level of sales of McClure’s in different districts.A 1-standard-deviation increase in the circulation of McClure’s in the districtleads to a 4-percentage-point increase in the probability of a pro-consumer vote(a 9 percent increase with respect to the sample mean).

A 1-standard-deviation change in McClure’s circulation corresponds to 1,187people. Can 1,187 additional informed people have such impact? According toCampbell and Jurek (2003, table 2), in 1900–1924 the median victory spread in




Table 3

Impact of Muckraking on House Votes on Domestic Regulatory Legislation, 1902–17

(1) (2) (3)

Muckraked # McClure’s Circulation 7.405** 7.572*(2.506) (3.568)

Muckraked .0296* .252**(.0120) (.0800)

McClure’s Circulation/population (district) 5.388** .719 .464(1.364) (1.708) (3.098)

x-Coordinate �.897** �1.177** �1.362**(.0641) (.129) (.170)

y-Coordinate .0286� .0101 �.00169(.0166) (.0328) (.0464)

Republican �.0535 .247* .188(.0522) (.108) (.152)

Third party .0628 .0222 �.0348(.0780) (.243) (.175)

Muckraked # controls N.A. Yes Yes

Note. The sample is based on 40 final roll call votes. The dependent variable is a dummy variable thattakes the value of one if it is a consumer-oriented vote and zero otherwise. All regressions include controlsfor a senator’s x- and y-coordinates from Poole and Rosenthal (1997, 2001), Republican and third-partydummies (with Democrat the excluded category), and nine census division fixed effects. The coefficientsreported are the marginal effect of a probit estimation computed at the average value of the independentvariables. The standard errors (in parentheses) are heteroskedasticity robust and clustered at the congres-sional district level. . N.A. p not applicable.N p 7,497

� p ! .1.* p ! .05.** p ! .01.

a congressional margin was 11.8 percent. Since the average district had a pop-ulation of 193,200 and only 15 percent of people voted (based on presidentialelections), the median congressional election was decided by 1,700 people. Fitz-patrick (1994) estimates that at that time muckraking magazines’ readership wasup to six times their circulation. Thus, a 1-standard-deviation increase in thecirculation of a muckraking magazine implied that 7,122 more people in acongressional district were informed about an issue. Even assuming that half ofthem were women, who did not vote at that time, and only half of the malereaders’ voting preferences shifted as a result of the information contained inthe muckraking magazine, a 1-standard-deviation change in circulation wouldpotentially be able to move 1,780 voters, a sufficient number of voters to changethe results in at least half of the congressional elections.

Finally, in column 3 we include an issue fixed effect. If there are topics whereeveryone tends to vote more pro-consumer, this should be captured by thesefixed effects. The results are unchanged.

3.6. Evidence from Senate Votes on Regulatory Legislation

This result does not necessarily imply that the treatment of certain issues bythe muckraking magazines led congressmen to alter their votes. An alternativeinterpretation of our results is that the media, instead of catering to the audience’s




demand for entertainment, cater to the demand for information (Gentzkow andShapiro 2006). In this case, districts where voters are more sensitive to certainissues are likely to have both higher diffusion of magazines that cover thoseissues and representatives who are more likely to vote accordingly (in responseto their voters’ exogenous preferences). This interpretation is able to accountfor most of our empirical results, without assuming any causality between news-paper reporting and congressional votes.

To assess the strength of this alternative interpretation, in Table 4 we test thesame specifications of the determinants of voting behavior on regulatory legis-lation applied to members of the Senate. Until 1913, most senators were notelected but were appointed by state governors. Therefore, they reflected theprevailing ideology in the state but were not necessarily sensitive to voters’pressure. If muckraking magazine circulation is just a proxy for the componentof the prevailing ideology in a certain area that is not captured by the Poole andRosenthal coordinates, then circulation should also affect the way senators voted,since they should share the same ideology. By contrast, if elected officials simplyreact to the voters’ awareness created by the muckraking articles, then the votesof appointed senators should not necessarily be affected by muckraking magazinecirculation.

In Table 4, we reestimate the specification from Table 3 using the Senate’sregulatory votes with standard errors clustered at the state level. As columns 1–3 show, the diffusion of McClure’s in the state did not increase the probabilityof a pro-consumer vote. This measure has a negative effect in all columns,sometimes statistically significant. The interaction of circulation with the issuesthat are muckraked has a positive coefficient, but it is never statistically signif-icant. This result is consistent with the hypothesis that McClure’s circulation hadno effect on the probability of a pro-consumer vote by nonelected officials.

The best evidence of the differential response of elected and nonelected of-ficials, however, can be obtained by looking at changes in behavior in the Senateas more senators were elected. Even before the introduction of the SeventeenthAmendment, there was a movement at the state level to increase the account-ability of senators. Oregon was the first state to introduce provisions in the statelegislature and constitution (later collectively called the Oregon plan) to ensurethat its U.S. senators were directly elected rather than appointed. By 1911, 20states had adopted similar plans, and a variety of other states had taken smallersteps in the same direction.

Did senators who expected to face an election behave in a different way? Totest this hypothesis, Table 4 also splits the sample between appointed senatorsand senators from states with a provision for direct election by 1911. The in-teraction between muckraked issues and circulation of muckraking journals ispositive and statistically significant only among elected senators. This evidenceillustrates the mechanism through which this influence takes place. Elected of-ficials are sensitive to informed voters, and media help voters become informedeven on issues where their direct interest in being informed is minimal.



Tab

le4

Impa

ctof

Mu

ckra

kin

gon

Sen

ate

Vot

eson

Dom

esti

cR

egu

lato

ryLe

gisl

atio

n,

1902

–17

(1)

(2)

(3)

(4)

(5)

Mu

ckra

ked

#M

cClu

re’s

Cir

cula

tion

(sta

te)

31.9

934

.52

�.5

6218

9.1*

*(2

7.85

)(3

6.94

)(4

5.55

)(6

1.74

)M

uck

rake

d�

.025

7�

.390

*�

.388

�

(.03

77)

(.16

2)(.

212)

McC

lure

’sC

ircu

lati

on/1

910

popu

lati

on(s

tate

)�

1.43

1�

28.8

2�

36.1

9�

23.3

7�

178.

9**

(7.6

81)

(21.

86)

(29.

50)

(35.

61)

(63.

34)

x-C

oord

inat

e�

.728

**�

1.16

5**

�1.

374*

*�

1.63

9**

�1.

291*

*(.

183)

(.26

5)(.

319)

(.39

2)(.

485)

y-C

oord

inat

e.0

455

�.0

355

�.0

471

�.4

25**

.029

6(.

0442

)(.

0803

)(.

0996

)(.

143)

(.15

7)R

epu

blic

an�

.316

*�

.178

�.0

983

�.2

81�

.221

(.13

5)(.

184)

(.21

7)(.

257)

(.40

8)T

hir

dpa

rty

�.1

89�

.422

**�

.475

**�

.464

**(.

185)

(.06

59)

(.07

30)

(.11

9)Is

sue

fixe

def

fect

sN

oN

oYe

sYe

sYe

sSt

ates

incl

ude

dA

llA

llA

llA

ppoi

nte

dD

irec

tel

ecti

onM

uck

rake

d#

con

trol

sN

.A.

Yes

Yes

Yes

Yes

N2,

161

2,16

12,

161

1,23

892

3

Not

e.T

he

sam

ple

isba

sed

on34

fin

alro

llca

llvo

tes.

Th

ede

pen

den

tva

riab

leis

adu

mm

yva

riab

leth

atta

kes

the

valu

eof

one

ifit

isa

con

sum

er-o

rien

ted

vote

and

zero

oth

erw

ise.

All

regr

essi

ons

incl

ude

con

trol

sfo

ra

sen

ator

’sx-

and

y-co

ordi

nat

esfr

omP

oole

and

Ros

enth

al(1

997,

2001

),R

epu

blic

anan

dth

ird-

part

ydu

mm

ies

(wit

hD

emoc

rat

the

excl

ude

dca

tego

ry),

and

nin

ece

nsu

sdi

visi

onfi

xed

effe

cts.

Col

um

n4

use

sth

esa

mpl

eof

sen

ator

sfr

omst

ates

wit

hn

opr

ovis

ion

for

dire

ctel

ecti

on;c

olu

mn

5u

ses

the

sam

ple

ofst

ates

wit

ha

prov

isio

nfo

rdi

rect

elec

tion

asof

1911

.C

oeffi

cien

tsar

eth

em

argi

nal

effe

cts

ofa

prob

ites

tim

atio

nco

mpu

ted

atth

eav

erag

eva

lue

ofth

ein

depe

nde

nt

vari

able

s.H

eter

oske

dast

icit

y-ro

bust

stan

dard

erro

rsar

ein

pare

nth

eses

and

are

clu

ster

edat

the

stat

ele

vel.

N.A

.p

not

appl

icab

le.

�p

!.1

.*

p!

.05.

**p

!.0

1.




This test helps to alleviate the concern that our results are driven solely bythe fact that muckraking magazine circulation is a proxy for the prevailingideology in a certain area. Yet it is still possible—albeit unlikely—that our resultsare driven by the fact that muckraking magazine circulation is a proxy for somevoters’ preferences that are not captured by a senator’s party affiliation or byhis Poole and Rosenthal coordinates.

Stromberg (2004a, 2004b) shows that media are effective in informing inter-ested voters about the policies that benefit them directly. We show here that thepoint is more general and also applies to public interest policies where no voterhas any significant personal interest. While this distinction seems small, its im-plications are big. If media are effective only in informing interested voters, thenthey will cater to special interests, exacerbating the problem identified by theETR. If, by contrast, media are also effective in informing voters about issuesof public interest, then media can be a countervailing force to special interestsin the regulatory arena.

4. Evidence from Votes on the Seventeenth Amendment

Here we introduce an additional test as a further attempt to address thepotential competing interpretation for our findings that media attention is justcapturing demand for information.

By the dawn of the twentieth century, there was growing discomfort with theconstitutional provision that permitted U.S. senators to be appointed by theirstate governments rather than directly elected by their constituents. Some criticscharged that the insulation of the senate selection process from the will of thevoters allowed business interests to control senate appointments and pick sym-pathetic senators who were likely to support their special interests. As journalistCharles Edward Russell said, “Strictly speaking we had no Senate; we had onlya chamber of butlers for industrialists and financiers” (quoted in Phillips 1964,p. 20).

In line with this public concern, legislators took steps to amend the Consti-tution to mandate the direct election of U.S. senators. The ratification processrequired, first, that both the House and the Senate pass the amendment withtwo-thirds majorities and, next, that the amendment be approved by three-quarters of the states. While there was clear support in the House of Represen-tatives for such a move (with votes that were nearly unanimous in 1893, 1894,1898, 1900, 1902, and 1911), the Senate generally refused to bring the issue toa vote. When the Senate finally did allow for a roll call vote in 1902, the proposedamendment failed by a significant margin. Another 9 years passed before theSenate again voted on the amendment, but this time the provision passed. Afterthe requisite supermajority was achieved in the states, the Seventeenth Amend-ment was officially ratified in 1913.

Between the two votes on the proposed amendment, more precisely in 1906,




Figure 2. Diffusion of Cosmopolitan per 1,000 people in 1915

Cosmopolitan published the above-mentioned Treason of the Senate series, amajor expose of corruption in the Senate. To explore whether exposure to thestories in this series influenced senators’ voting patterns, we take advantage ofthe fact that we have information on the sales of Cosmopolitan by state. Thisinformation was difficult to assemble, as Cosmopolitan did not keep such records.Fortunately, in 1914 the Audit Bureau of Circulations was created to measurethe circulation of newspapers and magazines across the country. We contactedthe bureau and assembled the data for the first year available for Cosmopolitan(1915). We paired this information with data from the census on populationper state, which was available for 1910 and 1920, and we calculated the averagevalue for 1915. In the key regressions presented below, our measure of thediffusion of muckraking is the number of copies of Cosmopolitan sold in a statein 1915 divided by the state population in 1915 (in thousands of people).

Figure 2 illustrates the diffusion of Cosmopolitan by state, measured by thenumber of copies sold per thousand people. On average, Cosmopolitan sold 11copies per thousand inhabitants and McClure’s sold six (Figure 1). But there wasa wide dispersion. Both magazines were sold more in the West and in theNortheast, while they were less present in the South. As is evident in comparingFigure 1 with Figure 2, there was quite a large overlap between circulation ofMcClure’s and Cosmopolitan, with a correlation of .91.

To increase our confidence that we are estimating the effect of differences inthe availability of information about corruption in the Senate, it is importantto control for other factors that may have influenced senate voting behavior.Two relevant issues are the presence of provisions for direct election already at




Table 5

Determinants of Senators’ Voting Behavior on the SeventeenthAmendment: Summary Statistics

Variable N Mean SD Min Max

1911 Vote 91 .725 .449 0 1Switch 82 .329 .649 �1 1Cosmopolitan Circulation/population (1,000s) 91 10.539 4.903 2.530 26.368McClure’s Circulation/population (1,000s) 91 5.772 3.057 1.599 13.887x-Coordinate 91 .025 .441 �1.220 .708y-Coordinate 91 �.278 .640 �1.343 1.236Republican 91 .549 .500 0 1Direct Election 91 .440 .499 0 1Contested Selection (or Election) 91 .220 .416 0 1

Note. The term 1911 Vote is a dummy variable that takes the value of one if the senator voted in favorof the amendment. Switch is a variable that takes the value of �1, 0, or 1 depending on whether thesenator’s seat changed from a yes vote to a no vote between 1902 and 1911, remained unchanged, orchanged from a no vote to a yes vote. Direct Election is a dummy variable that takes the value of one ifa state had adopted a version of the Oregon plan. Contested Selection (or Election) takes the value of oneif selection (or election) of a senator was contested.

the state level, as mentioned earlier, and increased awareness of the U.S. Senateat the state level arising from other sources, such as a sharply contested ordisputed selection process.17 To capture these two issues, we include two dummyvariables: one that identifies whether the state had moved toward direct electionof senators on its own (that is, whether it adopted the Oregon plan) by 1911and another equal to one if the state experienced a contested selection prior to1911.

Table 5 presents the summary statistics for our 1911 sample of senators. Wehave data for 91 senators (of a total of 92 senators, or two from each of 46states, at the time). A slight majority of these senators were Republicans, and44 percent represented states with some provision for the direct election ofsenators.

In the first column of Table 6, we estimate a simple probit model of theprobability that a senator voted in favor of the Seventeenth Amendment as afunction of Poole and Rosenthal’s x- and y-coordinates and a Republican Partydummy (there were no third-party senators). As an additional control for regionaleffects on voting, in columns 1–4 we use the nine census division dummies.Some of these dummies perfectly predict the vote, so we lose 20 observations.For this reason, in columns 5–8 we reestimate with the six macroregion dummies(New England, North East, South East, North Central, South Central, and West).The coefficients reported are the marginal increase in the probability of a yesvote calculated at the average value of the independent variables. Right-wingsenators (a high x-coordinate) were less likely to vote in favor of the Seventeenth

17 A second important issue was that there might be heightened attention arising independent ofcorruption concerns. Notably, some disputes at the state level resulted in no senator being appointedat the requisite time. As a result, states could be underrepresented in the Senate for significantperiods.



Tab

le6

Impa

ctof

Mu

ckra

kin

gon

Sen

ator

s’19

11V

ote

onth

eSe

ven

teen

thA

men

dmen

t

Nin

eC

ensu

sD

ivis

ion

Du

mm

ies

(Np

71)

Six

Reg

ion

alD

um

mie

s(N

p91

)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Cos

mop

olit

anC

ircu

lati

on/p

opu

lati

on(s

tate

).0

467*

.043

7*.0

541*

.022

9�.0

200�

.024

0�

(.02

05)

(.01

94)

(.02

51)

(.01

36)

(.01

17)

(.01

43)

x-C

oord

inat

e�

1.31

5**

�1.

107*

*�

1.02

3*�

1.09

7**

�.6

43**

�.5

13*

�.4

63*

�.5

15*

(.32

1)(.

316)

(.43

2)(.

335)

(.23

7)(.

207)

(.20

6)(.

202)

y-C

oord

inat

e.4

19**

.297

*.2

26*

.281

*.1

82**

.137

*.1

16*

.136

*(.

139)

(.13

1)(.

105)

(.12

5)(.

0625

)(.

0631

)(.

0587

)(.

0633

)D

irec

tE

lect

ion

�.2

52*

�.0

681

(.11

7)(.

0518

)C

onte

sted

Sele

ctio

n(o

rE

lect

ion

).1

07.0

618�

(.08

98)

(.03

73)

McC

lure

’sC

ircu

lati

on/p

opu

lati

on(s

tate

)�

.018

7�

.002

63(.

0499

)(.

0216

)M

uck

rake

d#

con

trol

s.7

66**

.595

*.4

11�

.583

*.4

63�

.299

.189

.298

(.17

2)(.

251)

(.24

8)(.

265)

(.25

6)(.

245)

(.17

3)(.

245)

Not

e.C

oeffi

cien

tsar

efr

ompr

obit

regr

essi

ons

that

esti

mat

eth

eim

pact

ofth

epr

obab

ility

ofa

mar

gin

alch

ange

inth

eri

ght-

han

dva

riab

le,

calc

ula

ted

atth

em

ean

valu

e.H

eter

oske

dast

icit

y-ro

bust

stan

dard

erro

rsar

ein

pare

nth

eses

and

are

clu

ster

edat

the

stat

ele

vel.

Con

trol

sin

clu

dea

sen

ator

’sx-

and

y-co

ordi

nat

esfr

omP

oole

and

Ros

enth

al(1

997,

2001

),ge

ogra

phic

aldu

mm

ies,

and

the

Rep

ubl

ican

dum

my.

�p

!.1

.*

p!

.05.

**p

!.0

1.




Amendment, while senators with a y-coordinate with a higher value were morelikely to vote in favor. Interestingly, when we control for their x-coordinate,Republican senators were more likely to vote for the Seventeenth Amendment(that is, progressive Republicans were in favor).

In column 2 we include the circulation of Cosmopolitan. Senators from stateswhere Cosmopolitan was more diffused were more likely to vote in favor of theSeventeenth Amendment. A 1-standard-deviation increase in the diffusion ofCosmopolitan increased the probability of a yes vote by 35 percent. This effectpersists when we control for other possible determinants of the vote. In column3 we include a dummy equal to one if a senator came from a state with someprovision for direct voting and a dummy equal to one if a senator’s most recentselection was contested. The coefficient of the diffusion of Cosmopolitan isunchanged.

Even if we control for regional dummies, Cosmopolitan’s level of sales mightjust pick up any variable with a similar geographical concentration. It would beuseful to have a another magazine, with a similar pattern of concentration, thatdid not cover the Treason of the Senate, to use in the regression as a placebo.McClure’s is such a magazine. As Figures 1 and 2 show, the pattern of geographicaldiffusion is similar, but McClure’s sales have less reason to be causing the vote,because McClure’ s did not publish the Treason of the Senate series. In column4 we insert this variable. The effect of Cosmopolitan is substantially unchanged.

In columns 5–8, we reestimate the specifications with six regional dummies,which prevents us from losing 20 observations. The results are substantiallyunchanged, even if the magnitude of the coefficient dropped by half.

On the basis of these regressions alone, we cannot be sure that the effectcaptured by the diffusion of Cosmopolitan is not spurious. It is possible thatsenators from states where Cosmopolitan was very diffused were already naturallymore inclined to vote in favor of the Seventeenth Amendment, regardless of thepressure exerted by the Treason of the Senate series.

Fortunately, in 1902 senators voted on essentially the same amendment. Ifthe relationship visible in 1911 is spurious—that is, if senators from states withhigh diffusions of Cosmopolitan were more inclined to vote for the amendment,regardless of the muckraking articles that appeared in 1906—then one wouldexpect the same basic pattern of voting in 1902 as well. But this was not thecase. In unreported regressions, we found that the estimated effect of the diffusionof Cosmopolitan on the 1902 vote is often negative (not positive) and is neverstatistically significant.18

In Table 7, we look at changes in voting behavior between 1902 and 1911. Inthis way, we control for any state characteristics that did not change over time.In columns 1–3 we look at changes in the votes of senators from the same state,while in columns 4 and 5 we look at the change in vote of the same senator if

18 The regression was omitted to save space but is available from the authors on request.




Table 7

Impact of Muckraking on the Probability of a Change in Senators’ Votes on theSeventeenth Amendment between 1902 and 1911

By Senate Seat(N p 82)

By Same Senator(N p 20)

(1) (2) (3) (4) (5)

Cosmopolitan Circulation/population (state) .0737** .0809** .112* .206** .389*(.0275) (.0285) (.0490) (.0792) (.165)

Direct Election .445 �.765 1.788*(.299) (.523) (.870)

Contested Selection (or Election) �.0610 1.659** �.810(.323) (.487) (1.077)

x-Coordinate �2.864**(.901)

y-Coordinate .0760(.381)

Republican 1.442�

(.825)Constant .594� .364 .180 �.367 �1.886�

(.347) (.379) (1.061) (.751) (1.122)Census division fixed effects No No Yes No No

Note. The dependent variable Switch takes the value of �1, 0, or 1, depending on whether the senateseat changed from a yes vote to a no vote, remained unchanged, or changed from a no vote to a yes vote.Coefficients are from an ordered probit regression with robust standard errors in parentheses and clusteredat the state level.

� p ! .1.* p ! .05.** p ! .01.

he was not replaced in the interim. For this reason, the number of observationsin the final specification drops to 20.

We assign a value of �1 if the vote changed from no to yes, 0 if it did notchange, and �1 if it changed from yes to no. As a consequence, we estimate anordered probit regression.19 Column 1 reports results in which the only explan-atory variable for the change in voting behavior is exposure to the Treason ofthe Senate series, measured by the ratio of Cosmopolitan’s circulation to thestate’s population. The diffusion measure has a statistically and economicallysignificant impact on the probability of switching to a yes vote.

In column 2 we control for two other factors that also changed between 1902and 1911: the possibility that the state introduced provisions for direct electionand the possibility that the state experienced a contested selection of a senator.These controls create a slight increase in the economic and statistical significanceof the diffusion measure. In column 3 we introduce the additional control ofthe Poole and Rosenthal x- and y-coordinates in 1911 and a Republican Party

19 The standard ordered probit regression (see, for example, Stata) imposes the parallel regressionassumption, in which the effects of the explanatory variables do not vary with the point at whichthe categories of the dependent variable are dichotomized. Our results are robust to a more generalformulation, namely, the generalized ordered probit introduced by Maddala (1983) and Terza (1985).




dummy, producing substantially similar results. These controls allow for possibleideological changes between senators coming from the same state.

In columns 4 and 5, we restrict our attention to the 20 senators who werepresent in the Senate both in 1902 and in 1911. We reestimate the same modelwhere the dependent variable is change within senator and find similar resultsfor the importance of a muckraking magazine’s circulation on voting behavior.Indeed, the effect is even larger. Given that they are the same senators, there isno need to control for individual effects, and controlling for direct and contestedelections leads to larger coefficient estimates for diffusion in column 5.

In sum, the diffusion of Cosmopolitan in a state seems to have influenced itssenators’ position on the Seventeenth Amendment. One possible reason why allsenators, not just the elected ones, become more sensitive is that if the amend-ment was successful, they would all face scrutiny by the public. This effect doesnot seem to be driven by a spurious correlation between the diffusion of Cos-mopolitan and political preferences, since the diffusion of Cosmopolitan in asenator’s state does not have any predictive power on his 1902 vote on the sameissue, and the effect is present even when we control for the diffusion of a similarmagazine (McClure’s) that did not publish the Treason of the Senate series. Moretelling, the probability that a senator changed his vote between 1902 and 1911is correlated with the diffusion of Cosmopolitan in his state. It is difficult toexplain both the House and Senate evidence on regulatory issues and the changein the Senate vote on the Seventeenth Amendment without media having a rolein influencing the voting behavior of elected representatives.

5. Implications for the Economic Theory of Regulation

During the muckraking period, the media was powerful when it came topromoting regulation. This does not imply that the media always have this impactnor that all the regulation passed under the media pressure is beneficial toconsumers. The natural question, then, is when are media more or less likelyto play this role as a countervailing force? Answering this question provides apositive theory of regulation. Our historical examination suggests that threeadditional factors need to be considered.

First, both the interest and the ability of media to inform voters about anissue are directly linked to the newsworthiness of that issue. Some issues, suchas the safety of drinking water, are by their nature more conducive to entertainingstories (for example, the movie Erin Brockovich). For this type of issue, mediacoverage can more easily shift the balance of power in favor of broad, dispersedinterests. In fact, it can even drive politicians to take up populist initiatives. Yetthe implementation of regulation is much less newsworthy, which explains whyCongress can be pushed by public opinion to approve pro-consumer legislationthat later, in the implementation phase, may be wholly or partly captured byvested interests.




Although it is difficult to say with precision what makes an issue newsworthy,two likely factors are when it is novel and when it seems out of the ordinary.With respect to the first factor, regulation that serves broad and dispersed interestsis more likely to be enacted when the underlying issue is novel or when a newmeans of communication puts an old issue in a new light (for example, as therise of television did for natural disasters). Indeed, it may be more than acoincidence that each of the three major eras of policy reform in the twentiethcentury (the Progressive Era, the New Deal, and the Great Society) immediatelyfollowed the introduction of a new technology for mass communication (nationalperiodicals, radio, and television, respectively). Similarly, news that is out of theordinary is more likely to attract attention than coverage that is familiar. Thismay help to explain why public demand for regulation sometimes seems to bedirected toward addressing the exception, rather than the norm, including verylow probability events such as shark attacks (Sunstein and Zeckhauser 2010).

More broadly, in the framework we have used in this paper, the media aremotivated to convey information to the public out of a desire to maximizeprofits. This is the likely goal of media whose ownership is dispersed or whoselarge shareholders own only (or principally) media companies. By contrast, whencontrolling shareholders in media companies have a vested interest in industry(for example, when they also own regulated firms outside of the media sector)or if they have a specific political goal, the media might pursue a different agenda.Where media are concentrated in the hands of industrial interests, their populistimpulse (that is, to appeal to a broad group of consumers) may be outweighedby the particular political interests of their owners (for example, to avoid reg-ulation of their industry). In such a situation, not only can the media lose theirbeneficial role, but they may actually serve or become part of the so-called factoryof consensus dreaded by Herman and Chomsky (1998). If we assume that do-mestic owners of media outlets are more likely than foreign ones to have politicalobjectives (beyond pure profit maximization), this observation can help to ex-plain why corruption is negatively correlated with foreign ownership of the media(Besley and Prat 2006) and with government ownership of the media (Djankovet al. 2003).20

Third, we have assumed that profit maximization derives from sales maxi-mization, while the most important source of revenues for modern media isadvertising, not sales. Interestingly, early muckraking magazines relied almostexclusively on sales (which may help to explain their activism). Advertisingrevenues very often follow circulation, and the two goals coincide. Yet to anadvertiser, the value of an audience is driven by its aggregate purchasing power.Hence, smaller, more wealthy groups can receive more media attention thanlarger but less wealthy ones. This suggests that even when the media play a large

20 Of course, not all foreign owners of media outlets are devoid of political objectives, as criticsof Australian-born Rupert Murdoch (and his Fox network in the United States) regularly make clear.Murdoch became a naturalized citizen of the United States in 1985.




role in presenting information on particular issues and thus blunt the influenceof large incumbents, they may still cater to the interests of a relatively affluentconstituency.

Sales maximization also can diverge from profit maximization when advertisershave some market power over the media outlet. Reuter and Zitzewitz (2006),for example, find that financial advice is biased in favor of advertisers inpublications with a more concentrated set of advertisers. The idea is that in thesepublications, the power of a media outlet to resist the pressure of each advertiseris reduced. Having a more concentrated set of advertisers may thus lead to aless inquisitive press and, in turn, to a less informed electorate and, thus, topolitical decision making that is more responsive to private interests. Regardlessof a country’s internal industrial structure, its set of potential advertisers isgenerally larger if its economy is open. Hence, the opening up of an economymay provide an additional benefit by reducing advertisers’ power vis-a-vis themedia and thus freeing the media to inform the public.

All these factors help to explain why muckraking arose, flourished, and sub-sided when it did. As Glaeser and Goldin (2006) argue, at the beginning of thetwentieth century, major technological innovations made magazine productiondramatically cheaper, which opened up a large market for periodicals, a marketmade even larger by the increasing level of literacy. The muckraking magazines,initially literary publications, discovered that the marketing power of investigativejournalism was driving their remarkable and rapid increases in circulation. Themuckraking magazines’ success, however, reduced the newsworthiness of addi-tional inquiries by saturating the public with news of scandals. Their success inbuilding circulation also led to increased opportunities to advertise and thusincreased influence of advertisers, who were likely less than enthusiastic aboutinvestigative pieces critical of business.

If profit-maximizing media have the potential to inform the public and thusto limit the power of special interests, then the cycle just recounted—a cycle ofnew communications technology, followed by vigorous investigative journalismto build market share, and finally by retreat from aggressive investigation—mayhelp to explain why major reform initiatives tend to be episodic and why (inthe United States at least) the key reform eras of the twentieth century (Pro-gressive Era, New Deal, and Great Society) tended to follow the introduction ofmajor new communications technologies (mass printing, radio, and television,respectively).

6. Conclusion

The economic theory of regulation emphasizes the power of narrow, concen-trated interests over broad, dispersed ones. In this paper, we argue that a missingelement from this theory is the role of profit-maximizing media. In representativedemocracies, profit-seeking media increase the relative power of broad, highly




dispersed groups by reducing their cost of becoming informed. Motivated toreach big audiences by the lure of large profits, media firms typically seek totransform real events and issues—including public policy issues—into enter-taining stories. In so doing, they end up informing the public about these issuesand events, thus helping to overcome the problem of rational ignorance high-lighted by Downs (1957). By informing voters, media help to make electedrepresentatives more sensitive to the interests of their constituencies and lessprone to excessive influence or capture by special interests. Several characteristicsof the media market help determine to what extent the media will be a coun-tervailing force to private interests and to what extent they themselves will becaptured.

We document the importance of this channel by studying the impact thatmuckraking articles had on the voting behavior of U.S. representatives andsenators at the beginning of the twentieth century. We find this effect to be notonly statistically significant but also quantitatively large.

The results point toward a revised economic theory of regulation able todifferentiate where and when vested interests will prevail. Our analysis suggeststhat vested interests will have less influence over the legislative agenda on issuesthat are more newsworthy. They will also have less influence when media own-ership and advertising budgets are less concentrated.

References

Adams, Samuel Hopkins. 1905a. The Great American Fraud. I. Collier’s Weekly, October7, pp. 14–15, 29.

———. 1905b. The Great American Fraud. II: Peruna and the Bracers. Collier’s Weekly,October 28, pp. 17–19.

———. 1905c. The Great American Fraud. III: Liquozone. Collier’s Weekly, November18, pp. 20–21.

———. 1905d. The Great American Fraud. IV: The Subtle Poisons. Collier’s Weekly,December 2, pp. 16–18.

———. 1906a. The Great American Fraud. V: Preying on the Incurables. Collier’s Weekly,January 13, pp. 18–20.

———. 1906b. The Great American Fraud. VI: The Fundamental Fakes. Collier’s Weekly,February 17, pp. 22–24, 26, 28.

———. 1906c. The Great American Fraud. Series II: Quacks and Quackery. I: The Sure-Cure School. Collier’s Weekly, July 14, pp. 12–14, 22.

———. 1906d. The Great American Fraud. Series II: Quacks and Quackery. II: TheMiracle Workers. Collier’s Weekly, August 4, pp. 14–16.

———. 1906e. The Great American Fraud. Series II: Quacks and Quackery. III: TheSpecialist Humbug. Collier’s Weekly, September 1, pp. 16–18.

———. 1906f. The Great American Fraud. Series II: Quacks and Quackery. IV: TheScavengers. Collier’s Weekly, September 22, 16–18, 24.

Allen, Howard W., and Jerome Clubb. 1974. Progressive Reform and the Political System.Pacific Northwest Quarterly 65:130–45.




Baron, David. 1994. Electoral Competition with Informed and Uniformed Voters. Amer-ican Political Science Review 88:33–47.

Becker, Gary S., and Kevin M. Murphy. 1993. A Simple Theory of Advertising as a Goodor Bad. Quarterly Journal of Economics 108:941–64.

Besley, Timothy, and Ron Burgess. 2002. The Political Economy of Government Re-sponsiveness: Theory and Evidence from India. Quarterly Journal of Economics 117:1415–52.

Besley, Timothy, and Andrea Prat. 2006. Handcuffs for the Grabbing Hand? Media Captureand Government Accountability. American Economic Review 96:720–36.

Campbell, James E., and Steve J. Jurek. 2003. The Decline of Competition and Changein Congressional Elections. Pp. 43–72 in Congress Responds to the Twentieth Century,edited by Sunil Ahuja and Robert Dewhirst. Columbus: Ohio State University Press.

DellaVigna, Stefano, and Ethan Kaplan. 2007. The Fox News Effect: Media Bias andVoting. Quarterly Journal of Economics 122:1187–1234.

———. 2008. The Political Impact of Media Bias. Pp. 79–105 in Fact Finder, Fact Filter:How Media Reporting Affects Public Policy, edited by Roumeen Islam. Washington, D.C.:World Bank Publications.

DeNevi, Donald P., and Helen M. Friend. 1973. Muckrakers and Robber Barons. Danville,Calif.: Replica Books.

Djankov, Simeon, Claralee McLiesh, Tatiana Nenova, and Andrei Shleifer. 2003. WhoOwns the Media? Journal of Law and Economics 46:341–82.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper & Brothers.Dyck, Alexander, David Moss, and Luigi Zingales. 2008. Special Interests versus the Media.

Working Paper No. 14360. National Bureau of Economic Research, Cambridge, Mass.Eisensee, Thomas, and David Stromberg. 2007. News Floods, News Droughts, and U.S.

Disaster Relief. Quarterly Journal of Economics 122:693–728.Fitzpatrick, Ellen. 1994. Muckraking: Three Landmark Articles. New York: Bedford Books.Gentzkow, Matthew, and Jesse Shapiro. 2006. Media Bias and Reputation. Journal of

Political Economy 114:280–316.George, Lisa, and Joel Waldfogel. 2006. The New York Times and the Market for Local

Newspapers. American Economic Review 96:435–47.Glaeser, Edward, and Claudia Goldin. 2006. Corruption and Reform: Lessons from America’s

Economic History. Chicago: University of Chicago Press.Grossman, Gene M., and Elhanan Helpman. 1996. Electoral Competition and Special

Interest Politics. Review of Economic Studies 63:265–86.Hamilton, James T. 2004. All the News That’s Fit to Sell. Princeton, N.J.: Princeton Uni-

versity Press,.Herman, Edward, and Noam Chomsky. 1998. Manufacturing Consent: The Political Econ-

omy of the Mass Media. New York: Pantheon Books.Ireland, Alleyne. 1914. Joseph Pulitzer: Reminiscences of a Secretary. New York: Mitchell

Kennerley.Law, Marc, and Gary D. Libecap. 2006. The Determinants of Progressive Era Reform:

The Pure Food and Drugs Act of 1906. Pp. 319–42 in Corruption and Reform: Lessonsfrom America’s Economic History, edited by Edward L. Glaeser and Claudia Goldin.Chicago: University of Chicago Press.

Maddala, G. S. 1983. Limited-Dependent and Qualitative Variables in Econometrics. NewYork: Cambridge University Press.




Maris, Kenneth C. 1982. The Historical Atlas of United States Congressional Districts, 1789–1983. New York: Free Press.

Mathews, John L. 1908. Water Power and the Price of Bread. Hampton’s, October, pp.1–14.

———. 1909. Water Power and the Pork Barrel. Hampton’s, July, pp. 455–69.McClure Publications. 1917. Analysis of Distribution of Circulation: Based on the Issue

of May, 1917: McClure’s Magazine. New York: The McClure Publications, Inc.McDonagh, Eileen Lorenzi. 1992. Representative Democracy and State Building in the

Progressive Era. American Political Science Review 86:938–50.Moss, David A., and Mary Oey. 2010. The Paranoid Style in the Study of American

Politics. Pp. 256–98 in Government and Markets: Toward a New Theory of Regulation,edited by Edward J. Balleisen and David A. Moss. Cambridge: Cambridge UniversityPress.

Olson, Mancur. 1965. The Logic of Collective Action: Public Goods and the Theory of Groups.Cambridge, Mass.: Harvard University Press.

Peltzman, Sam. 1976. Toward a More General Theory of Regulation. Journal of Law andEconomics 19:211–40.

Phillips, David Graham. 1906a. The Treason of the Senate. II: Aldrich, the Head of ItAll. Cosmopolitan, April, pp. 628–38.

———. 1906b. The Treason of the Senate. V: Bailey, the “Patriot.” Cosmopolitan, July,pp. 267–76.

———. 1906c. The Treason of the Senate. IV: Chief Spokesman of “The Merger.” Cos-mopolitan, June, pp. 123–32.

———. 1906d. The Treason of the Senate. VI: Confusing the People. Cosmopolitan,August, pp. 368–77.

———. 1906e. The Treason of the Senate. IX: Fairbanks, Hale, and Frye. Cosmopolitan,November, pp. 77–84.

———. 1906f. The Treason of the Senate. III: Left Arm of the Monster. Cosmopolitan,May, pp. 3–12.

———. 1906g. The Treason of the Senate. I: New York’s Misrepresentatives. Cosmopolitan,March, pp. 487–502.

———. 1906h. The Treason of the Senate. VII: The Rise of Foraker. Cosmopolitan, No-vember, pp. 525–35.

———. 1906i. The Treason of the Senate. VIII: Thrifty Patriot Allison. Cosmopolitan,October, pp. 627–36.

———. 1964. The Treason of the Senate, edited by George E. Mowry and Judson A.Grenier. Chicago: Quadrangle Books.

Poole, Keith T. 2004. Voteview (computer file). Version 3.03c. University of Georgia,Department of Political Science, Athens. http://www.voteview.com.

Poole, Keith T., and Howard Rosenthal. 1997. Congress: A Political-Economic History ofRoll Call Voting. New York: Oxford University Press.

———. 2001. D-Nominate after 10 Years. Legislative Studies Quarterly 26:5–29.Prat, Andrea, and David Stromberg, 2013. The Political Economy of Mass Media. Pp.

135–87 in Applied Economics, vol. 2 of Advances in Economics and Econometrics. Cam-bridge, Cambridge University Press.

Reuter, Jonathan, and Eric Zitzewitz. 2006. Do Ads Influence Editors? Advertising andBias in the Financial Media. Quarterly Journal of Economics 121:197–227.


http://www.voteview.com



Romero, Francine Sanders. 2007. The Impact of Direct Election on Reform Votes in theU.S. Senate. Social Science Quarterly 88:816–29.

Sanders, Elizabeth. 1999. Roots of Reform: Farmers, Workers, and the American State, 1877–1917. Chicago: University of Chicago Press.

Sinclair, Upton. 1906a. The Condemned-Meat Industry. Everybody’s, May, pp. 608–16.———. 1906b. Stockyard Secrets. Collier’s, March 24, p. 24.Sinclair, Upton, and L. L. Seaman. 1905. Is Chicago Meat Clean? Collier’s, April 22, pp.

13–14.Steffens, Lincoln. 1910. It: An Exposition of the Sovereign Political Power of Organized

Business. Everybody’s 23:391–8, 449–60, 646–56, 813–25.———. 1911. It: An Exposition of the Sovereign Political Power of Organized Business.

Everybody’s 24:217–29, 455–66.Stigler, G. 1971. Theory of Economic Regulation. Bell Journal of Economics 2:3–21.Stromberg, David. 2004a. Mass Media Competition, Political Competition, and Public

Policy. Review of Economic Studies 71:265–84.———. 2004b. Radio’s Impact on Public Spending. Quarterly Journal of Economics 119:

189–221.Sullivan, Mark. 1905. The Patent Medicine Conspiracy against the Freedom of the Press.

Colliers, November 4, pp. 13–15, 25.Sunstein, Cass R., and Richard Zeckhauser. 2010. Dreadful Possibilities, Neglected Prob-

abilities. Pp. 116–23 in The Irrational Economist: Making Decisions in a Dangerous World,edited by Erwann Michel-Kerjan and Paul Slovic. New York: PublicAffairs Books.

Terza, J. 1985. Ordered Probit: A Generalization. Communications in Statistics—Theoryand Methods 14:1–11.

U.S. Bureau of the Census. 1975. Historical Statistics of the United States. New York:Cambridge University Press.

Weinberg, Arthur, and Lila Weinberg. 1961. The Muckrakers. New York: Simon & Schuster.



The University of Chicago The Booth School of … · The University of Chicago ... Introduction...

Documents

Transcript of The University of Chicago The Booth School of … · The University of Chicago ... Introduction...